International Journal of Advancements in Research & Technology, Volume 3, Issue 5, May-2014 ISSN 2278-7763 189 Detection of Suspicious URLs through Vision Techniques in Twitter Stream Prof. Jagadish.P, Prof. Anand.R, Nikitha.R, Abhilasha.A.R Dept of Computer Science BMS Institute of Technology ABSTRACT - The primary intention of 1. INTRODUCTION WARNINGBIRD is to detect the suspicious URLs through correlated redirect chain methodology. It will examine the correlated URL redirect chain and tweet context Twitter is an online social networking site that enables users to send and read short 140 characters text messages called tweets. IJOART information to detect the suspicious URLs. Registered users can read and post tweets but unregistered users can only read them. The major goal of the WARNINGBIRD is to But unfortunately, the problem is when detect the suspicious URLs. Suspicious the attackers send different individual URLs are nothing but the doubtful URLs suspicious URLs it becomes inefficient to which identify the suspicious URLs through the Malicious correlated redirect chain methodology as malwares, phishing etc. Conventional twitter in this method it starts identifying the suspicious URL detection system is based suspicious URLs through the common on URLs which were frequently shared. methodology. It detects the suspicious URLs In this paper, we propose 3 approaches for detecting suspicious URLs. Our first approach is visual content matching technique, second approach is based on the creation and delivery details of URLs and the third approach is through MAUDE model. Copyright © 2014 SciResPub. contains malicious elements correlated URL elements. include redirect viruses, chain which were frequently shared. It will examine the correlated URL redirect chain and tweet context information to detect the suspicious URLs. But unfortunately, the problem is when the attackers send different individual suspicious URLs it becomes inefficient to implement correlated redirect IJOART International Journal of Advancements in Research & Technology, Volume 3, Issue 5, May-2014 ISSN 2278-7763 190 chain to identify the suspicious URLs as in Authentic User Detection) when a recipient this the receive a tweet, the incoming twitter server suspicious URLs through the common will contact the alleged outgoing twitter URLs which were frequently shared. server to verify that it sent that specific method it starts identifying In this paper, we further improve the efficiency of detecting the suspicious URLs in twitter stream by providing 3 approaches. tweet. The remainder of the paper is organized as follows. Section 2 shows the related work on twitter spam detection. Section 3 shows a We make following contribution to this suspicious URL detection system through paper:- vision and other techniques. 1. 2. RELATED WORK We propose our first approach visual this In some work, the twitter suspicious URL approach, we consider some of the features detection systems were based on account what recognize features. Account features include the twitter duplication of an original page. To do this account creation date, no of followers and we analyze a webpage based on some of its friends and ratio of tweets containing URLs. content matching the technique. In IJOART human does to characteristics and based on the way it looks visually. Specifically, we record a number of characteristics including the page title text, number of links, images, forms, iframes, metatags and logo. By using this visual content matching technique it can prevent the web attacks like phishing etc. In another work, they proposed relation features based. The two important factors were distance and connectivity. They constructed a twitter graph based on these two factors distance and connectivity. However, it allocated more space and took much time to detect the suspicious URLs. 2. Our second approach is based on creation of the URL and delivery details of that URL In another approach, they proposed message tweet. Here we are finding the suspicious features based. It was analyzing the lexical URL based on the date and time the URL features of the tweets like length of the tweet was created and its delivery details to whom and the content of the tweets. all that URL was delivered. 3. Our third approach is we are developing a model called Copyright © 2014 SciResPub. MAUDE(Multi server IJOART International Journal of Advancements in Research & Technology, Volume 3, Issue 5, May-2014 ISSN 2278-7763 191 even if are image is resized or rotated. To 3. SUSPICIOUS URL identify the key points we use SIFT (Scale- DETECTION SYSTEM invariant feature transform) algorithm. Each THROUGH VISION AND key OTHER TECHNIQUES dimensional vector. By using this algorithm point is represented as multi- we can obstruct phishing and other web Twitter is an online social networking site. attacks. User can send and receive the messages. Messages are called as tweets. The attacker may send a tweet which contains the suspicious URL which can cause disaster to the system. To identify the suspicious URLs we are proposing three approaches they are:- SIFT algorithm extracts the key points from an image. These key points are invariant to affine transformation, scaling and illumination. During matching, the key points of two images are matched. If the percentage of matched key points is greater IJOART a. Visual content matching technique. b. Based on creation and delivery than a threshold then that page is considered as proper web page; otherwise it is a fake details of the URL. c. Developing a MAUDE model. Visual content matching technique web page. Based on creation and delivery details of the URL Our first approach is visual content matching technique. When a tweet occur we Our second approach relies on URL creation Analyze the visual contents like page title date and time. When the attackers send the text, number of links, images, forms, tweet containing suspicious URL. It detects iframes, metatags, and logo and obtains the based on the fact on when that URL was SSL (secure sockets layer) certificate of the created and to whom all it is delivered at URL. So by these factors it detects the that date and time. Based on these factors it suspicious URLs. detects the suspicious URLs. Page title text, number of links, images, Developing a MAUDE model forms, iframes, metatags, and logo are used as key points of a web page. These key points are robust features that can be kept Copyright © 2014 SciResPub. Our third approach is we have built a model called MAUDE (Multi server Authentic IJOART International Journal of Advancements in Research & Technology, Volume 3, Issue 5, May-2014 ISSN 2278-7763 192 User Detection).The basic idea of MAUDE Twitter Stream”-IEEE TRANSACTIONS model is when a recipient’s twitter server ON receives a tweet; the incoming twitter server COMPUTING, will contact the alleged outgoing twitter JANUARY 2013. server to verify that it sent that specific [2] tweet. If so, the tweet is received and Butkiewicz, Harsha V. Madhyastha, processed; otherwise, the tweet is suspicious and appropriate action is taken e.g. tweet is thrown into some folder with an explicit DEPENDABLE AND VOL. Indrajeet X, Singh, Srikanth,V. SECURE NO. Y, Michael Krishnamurthy, SateeshAddepalli “Twitsper: Tweeting Privately” IEEE 2013. alert or even deleted. [3] H. Kwak, C. Lee, H. Park, and S. Moon, 4. CONCLUSION “What is Twitter, a social network or a news media?” in Proc. WWW, 2010. IJOART We addressed and specified the problems of correlated redirect chain methodology in twitter suspicious URL detection system. We have discussed the groundwork for a method and tool that can detect suspicious URLs. A new method and approach of [4] G. Stringhini, C. Kruegel, and G. Vigna, “Detecting spammers on social networks,” in Proc. ACSAC, 2010. [5] J. Kiss, “Twitter Reveals It Has 100m Active Users,” Guardian, 8 Sept. 2011. detecting suspicious URLs in twitter is proposed to prevent the system from web attacks and to have a safer communication with enhanced security. And we also gained critical insight into how to effectively and efficiently find suspicious URLs. 5. REFERENCES [1] Sangho Lee, Student Member, IEEE, and Jong Kim, “WARNINGBIRD: Member, A Near IEEE Real-time Detection System for Suspicious URLs in Copyright © 2014 SciResPub. IJOART